# [OPE-L:5626] ope-l: Deer-hunting

andrew kliman (Andrew_Kliman@CLASSIC.MSN.COM)
Tue, 21 Oct 97 23:44:42 UT

[ show plain text ]

In response to the PIAF:

----------
From: owner-ope-l@galaxy.csuchico.edu on behalf of Allin Cottrell
Sent: Tuesday, October 21, 1997 3:21 PM
To: ope-l@galaxy.csuchico.edu
Subject: [OPE-L:5626] Re: ope-l: RE: In defence of correlation

Allin wrote: "That there remains systematic variation in the residuals of a
regression of Y on X does not make X a biased predictor of
Y, as statisticians use the term."

I'm not exactly sure what Allin means by this, but my point was that a
sector's aggregate value differs from its aggregate price, even on average.
Thus, I think the value is what statisticians call a biased estimator of the
price. Here's the way Wonnacott and Wonnacott (_Introductory Statistics_, pp.
148-49) put it:

"An unbiased estimator is one that is, *on the average*, right on target ...
Formally, we state the definition,

theta-hat is an unbiased estimator of theta if E(theta-hat) = theta.

... Of course, an estimator theta-hat is called biased if E(theta-hat) is
different from theta; in fact, bias is defined as this difference,

Bias = E(theta-hat) minus theta."

They have an accompanying graph of a biased estimator, theta-hat. Theta-hat
is normally distributed, with a mean that is higher than the "true theta." In
parallel manner, if you were to take a series of observations of a particular
sector's aggregate prices and values, and re-norm them so that the aggregate
price always equaled (say) 1, you'd get a distribution of values with a mean
different from 1, often very different.

Whatever terminology one uses, I think that makes values undesirable as
predictors of prices.

BTW, Paul's claim that the economy-wide price-value ratio = 1 reminds me of